Solving LED Dimming Issue: Edward's Conundrum

[shadow_tek001]

Hello folks. My first post on this Forum:

I have a conundrum, I have built a LED light fixture in my living room. The issue I have is that the LEDS have dimmed over the last 5 months by about 50%. I did not notice it until I had to replace one the other day, and it was way brighter then the existing ones. (tried 2 new ones to be sure).
My setup is a simple one,
110/120V AC input
Through a DC rectifier
into 42 LED's in series
Each LED is using about 2.7 volts.
The LED's were bought on EBAY, and are 5mm White LED 18000mcd 20mA 3.2-3.6 Forward Voltage (from China)

I thought this would be an ok setup, but as I mentioned, they are getting dimmer.
Any ideas why?



Thanks in advance!

Edward.

 

[Jiggy-Ninja]

Is there any sort of limiting resistor on these LEDs?

120 V_ac is actually the RMS value. When it's rectified, it has a peak of over 170 V. 42 LEDs in series with 3.5V forward voltage each only sums up to 147 V. If you don't have a current limiting resistor, the current surge is probably shortening their lifespan significantly.

Use a proper DC supply, the "brick" wall plugs that come with every electronic device, and current limiting resistors.

 

[jim hardy]

i think it is a good idea to use capacitance to set the current .

Large Xc in series with a bridge rectifier , LED's across bridge with resistors to force them to share current.

Dropping most of the voltage across Xc avoids those high current peaks ninja picked up on.
And it produces no heat.

use caps suitable for across-the-line, they are designed to withstand expected transients.

BTW el-cheapo LED's just do that "slow fade".. don't buy those $2 LED nite-lights at walmart they fade in a few months.

I had good luck with an outfit called "uperBrite Led's".

 

[shadow_tek001]

Is there any sort of limiting resistor on these LEDs?

120 V_ac is actually the RMS value. When it's rectified, it has a peak of over 170 V. 42 LEDs in series with 3.5V forward voltage each only sums up to 147 V. If you don't have a current limiting resistor, the current surge is probably shortening their lifespan significantly.

Use a proper DC supply, the "brick" wall plugs that come with every electronic device, and current limiting resistors.

Could I just add 5 more LEDs into the series and call it a day then (170/3.6v = 47)? Run them at 2.5V, and how much of a difference in light output would there be in a 0.2 V drop.


Edward

 

[shadow_tek001]

i think it is a good idea to use capacitance to set the current .

Large Xc in series with a bridge rectifier , LED's across bridge with resistors to force them to share current.

Dropping most of the voltage across Xc avoids those high current peaks ninja picked up on.
And it produces no heat.

use caps suitable for across-the-line, they are designed to withstand expected transients.

BTW el-cheapo LED's just do that "slow fade".. don't buy those $2 LED nite-lights at walmart they fade in a few months.

I had good luck with an outfit called "uperBrite Led's".

I'm not sure what this means LOL. What's a large Xc? is that just a big capacitor?

Edward

 

[jim hardy]

yes,, Xc is exactly that, a capacitor. X is its impedance.

here's the trouble with just series-ing up LED's on AC:

they turn on rather abruptly when voltage across them reaches nominal, say 2.7 volts apiece.

So when your sinewave reaches 2.7 X 42 = 113.4 volts they start to conduct.
BUt the current through them is not linear with voltage, more like exponential.
So as the sinewave approaches its peak voltage your current rises well above the LEDs' rated current.

so the current flowing through your LED's is not just a rectified sinewave like the voltage, , but a series of very high amplitude short duration peaks.
It is difficult to calculate the RMS value of those short peaks but as general rule the narrower the peaks , the higher the ratio of rms to average.
Google terms 'RMS' and 'Crest Factor' ...

RMS is the heating value and those high current pulses may be hurting your LED's
That's what ninja was stating, i think...

 

[shadow_tek001]

yes,, Xc is exactly that, a capacitor. X is its impedance.

here's the trouble with just series-ing up LED's on AC:

they turn on rather abruptly when voltage across them reaches nominal, say 2.7 volts apiece.

So when your sinewave reaches 2.7 X 42 = 113.4 volts they start to conduct.
BUt the current through them is not linear with voltage, more like exponential.
So as the sinewave approaches its peak voltage your current rises well above the LEDs' rated current.

so the current flowing through your LED's is not just a rectified sinewave like the voltage, , but a series of very high amplitude short duration peaks.
It is difficult to calculate the RMS value of those short peaks but as general rule the narrower the peaks , the higher the ratio of rms to average.
Google terms 'RMS' and 'Crest Factor' ...

RMS is the heating value and those high current pulses may be hurting your LED's
That's what ninja was stating, i think...

OK, I think I'm beginning to understand.
So, how do I determine which type of capacitor to use? and where does said capacitor go?

Edward

 

[Jiggy-Ninja]

Would a capacitor be a good idea? I don't think the reactance model works when there's a diode in series with it. The cap will charge on the forward pulse, and the diodes will prevent it from discharging. Either the cap will charge once on the first pulse and not allow any more current through or the voltage will keep stepping up on each pulse until the voltage rating is exceeded and the capacitor is destroyed.

An inductor might work like you want, but a resistor is much cheaper and easier to understand.

 

[shadow_tek001]

Would a capacitor be a good idea? I don't think the reactance model works when there's a diode in series with it. The cap will charge on the forward pulse, and the diodes will prevent it from discharging. Either the cap will charge once on the first pulse and not allow any more current through or the voltage will keep stepping up on each pulse until the voltage rating is exceeded and the capacitor is destroyed.

An inductor might work like you want, but a resistor is much cheaper and easier to understand.

Ok then, now I don't know what to do.

Again though, would adding 5 or 6 LEDs to the series help?


Edward

 

[jim hardy]

""Would a capacitor be a good idea? I don't think the reactance model works when there's a diode in series with it. The cap will charge on the forward pulse, and the diodes will prevent it from discharging. Either the cap will charge once on the first pulse and not allow any more current through or the voltage will keep stepping up on each pulse until the voltage rating is exceeded and the capacitor is destroyed.""

use a full wave bridge and it'll be okay because current flows on bith half-cycles. Place a meg or ten across cap so it doesn't hold a charge when you unplug it from wall... disassemble a nightlight and you'll see that's how the commercial units are built. just size your cap for LED forward current.

look in the allied catalog for 'acrosss the line' caps, they're always rated 400V or more and carry UL approval.

 

[mdjensen22]

Would a capacitor be a good idea? I don't think the reactance model works when there's a diode in series with it. The cap will charge on the forward pulse, and the diodes will prevent it from discharging. Either the cap will charge once on the first pulse and not allow any more current through or the voltage will keep stepping up on each pulse until the voltage rating is exceeded and the capacitor is destroyed.

An inductor might work like you want, but a resistor is much cheaper and easier to understand.

Jiggy:
http://ww1.microchip.com/downloads/en/AppNotes/00954A.pdf

 

[jim hardy]

here's a hobbyist link with an article regarding your exact application.

http://www.discovercircuits.com/H-Corner/AC-Powered.htm


uses a capacitor to set current and a clever addition for safety - a metal oxide resistor to act as a fuse if things go awry...

note parallel-reverse connected LED's to avoid DC buildup on the cap. Id still place a few megs across the cap for bleed-down after power off.


look into caps -

http://www.google.com/url?sa=t&rct=...2NSeBQ&usg=AFQjCNE6AXh1FDEK6qIeqzobtAF8uKoddw

 

[vk6kro]

We had a similar discussion early this year.

We were discussing the possibility of using capacitors to limit the current to a series string of LEDs.
Someone who had done it commercially said that the scheme proved unreliable because the LEDs tended to blow up if the power was applied at the peak of the mains cycle.
Here is the post:
https://www.physicsforums.com/showpost.php?p=3065241&postcount=8

I suggest you just use a resistor.

If you rectify the 120 volt mains and put a filter capacitor of about 47 uF across the output, you will get a steady 170 volts.
Then the output of this can be taken to a 2.2 K, 1 watt resistor in series with 36 white LEDs.

You are already doing this, so you know how to do it safely, but anyone else should get help if they are not sure about using high voltages safely.

There should be a 100 K 1 watt resistor across the capacitor, for safety.

Incidentally, there are LED lamps on Ebay that are already set up for mains voltages. Some use single high powered LEDs and some use multiple smaller LEDs.
As an example:
http://www.ebay.com.au/itm/E27-108-LED-Light-Bulb-Lamp-Lighting-110V-Energy-Saving-/270806992456?pt=AU_Lighting_Fans&hash=item3f0d5ac648

 

[jim hardy]

Someone who had done it commercially said that the scheme proved unreliable because the LEDs tended to blow up if the power was applied at the peak of the mains cycle.

indeed that's the second function of the series resistor, to limit inrush.
Since a sinewave is statistically near peak voltage a lot more of the time than it is near zero voltage , most of your power-ons will cause a current spike.

arcsin(0.5)= thirty degrees, so a sinewave spends 2/3 of its time at > 50% , and over 1/4 of its time above 90%.

the 220 ohm in thread referenced is less effective than the 1k suggested at the hobby forum.. at peak:
170v/1kohm=0.17A spike;
170v/220ohm=0.77A spike

But i have no personal experience applying overcurrent spikes to LED's so can't tell you what's threshold for damage..
That thermistor seems an elegant solution.

old jim